Role of X11 and ubiquilin as In Vivo Regulators of the Amyloid Precursor Protein in Drosophila

The Amyloid Precursor Protein (APP) undergoes sequential proteolytic cleavages through the action of β- and γ-secretase, which result in the generation of toxic β-amyloid (Aβ) peptides and a C-terminal fragment consisting of the intracellular domain of APP (AICD). Mutations leading to increased APP levels or alterations in APP cleavage cause familial Alzheimer's disease (AD). Thus, identification of factors that regulate APP steady state levels and/or APP cleavage by γ-secretase is likely to provide insight into AD pathogenesis. Here, using transgenic flies that act as reporters for endogenous γ-secretase activity and/or APP levels (GAMAREP), and for the APP intracellular domain (AICDREP), we identified mutations in X11L and ubiquilin (ubqn) as genetic modifiers of APP. Human homologs of both X11L (X11/Mint) and Ubqn (UBQLN1) have been implicated in AD pathogenesis. In contrast to previous reports, we show that overexpression of X11L or human X11 does not alter γ-secretase cleavage of APP or Notch, another γ-secretase substrate. Instead, expression of either X11L or human X11 regulates APP at the level of the AICD, and this activity requires the phosphotyrosine binding (PTB) domain of X11. In contrast, Ubqn regulates the levels of APP: loss of ubqn function leads to a decrease in the steady state levels of APP, while increased ubqn expression results in an increase in APP levels. Ubqn physically binds to APP, an interaction that depends on its ubiquitin-associated (UBA) domain, suggesting that direct physical interactions may underlie Ubqn-dependent regulation of APP. Together, our studies identify X11L and Ubqn as in vivo regulators of APP. Since increased expression of X11 attenuates Aβ production and/or secretion in APP transgenic mice, but does not act on γ-secretase directly, X11 may represent an attractive therapeutic target for AD.


Introduction
One of the pathological hallmarks of Alzheimer's disease (AD) is the accumulation of amyloid plaques consisting of toxic b-amyloid (Ab) peptides. These peptides arise from the sequential cleavage of the Amyloid Precursor Protein (APP), a type I transmembrane protein, by two proteases known as band c-secretase (Fig. 1A). APP proteolysis by b-secretase generates an APP C-terminal fragment (CTF) known as C99. Subsequent cleavage of C99 by csecretase results in the release of Ab into the lumen and the APP intracellular domain (AICD) into the cytosol, where it can contribute to a transcriptional regulatory complex [1]. In addition to this amyloidogenic pathway, APP can also undergo nonamyloidogenic processing via sequential cleavage by aand csecretase (Fig. 1A). a-secretase cleaves within the Ab sequence, thereby precluding the formation of Ab. a-cleavage produces an APP CTF known as C83, which also serves as a substrate for csecretase activity [1].
c-secretase activity resides in a multi-protein complex that minimally contains Presenilins, Nicastrin, Aph-1 and Pen-2 [2,3]. Mutations in APP, Presenilin 1 and Presenilin 2 cause familial, early onset AD [4][5][6][7]. In addition, the triplication of the APP locus as well as promoter mutations in APP that increase APP expression are associated with AD [8,9]. The function of the AICD may also be crucial for AD pathogenesis, since each time Ab is generated, AICD is simultaneously released. AICD, in conjunction with two PTB domain-containing proteins (Fe65 and Tip60), can enter the nucleus and regulate the transcription of target genes, including APP itself [10]. In addition, AICD has also been implicated in other processes including cell signaling, apoptosis and calcium homeostasis [11][12][13][14][15][16][17][18][19][20][21]. Therefore, identifying genes that regulate APP steady-state levels, APP cleavage, and the fate and activity of AICD are likely to further our understanding of AD pathogenesis.
The X11/Mint protein family consists of three mammalian members: X11a and X11b, which are expressed in neurons, and X11c which is ubiquitously expressed. All three X11 proteins contain a phosphotyrosine binding (PTB) domain, followed by two PSD95/Dlg/ZO-1 (PDZ) domains [22]. Several observations suggest links between X11 and AD. First, X11a and X11b have been found in amyloid plaques in post-mortem AD brains [23,24]. Second, increased X11a and X11b expression in mammalian cells leads to a reduced secretion of extracellular Ab [25][26][27], while transgenic mice expressing either X11a or X11b are associated with reduced levels of Ab [28,29]. Third, X11 proteins physically interact with AICD via their PTB domains [30,31] and inhibit AICD-dependent transcription [32]. Fourth, X11a and X11b overexpression increases APP steady-state levels both in vitro and in vivo, likely due to altered maturation in the secretory pathway or endocytic trafficking of APP [25,27,33,34]. Finally, the X11 proteins bind to presenilin via their PDZ domains [35]. Furthermore, X11a and X11b have been reported to modulate c-secretase in mammalian cells [36]; however, alterations in the levels of C83 and C99 are not observed in transgenic mice overexpressing X11 [29]. These conflicting observations leave it unclear whether X11 overexpression can regulate c2secretase activity. The sequential actions of either a-secretase or b-secretase, which cleaves APP in its extracellular domain, and c-secretase, which cleaves APP within its transmembrane domain, generate AICD and C83 or C99, respectively. a-secretase cleaves APP within the sequence of Ab, thus precluding the generation of Ab. (B) GAMAREP (GMR-C99-Gal4, UAS-grim) contains two components: First, C99-Gal4, a chimeric protein with an N-terminal cleavable signal sequence, transmembrane and intracellular domain (yellow bar) and a C-terminal Gal4 (blue circle) is specifically expressed in the eye and is a substrate for c-secretase. Second, Grim (purple bar) is expressed under the control of a Gal4-dependent promoter (green bar) as a readout. In the presence of endogenous c-secretase activity (scissor), the unleashed AICD-Gal4 (yellow bar and blue circle fragment) translocates to the nucleus (dashed curve) and binds to the UAS element and activates (black arrow pointing to the right) Grim expression. This leads to apoptosis in the eye, resulting in flies with small and rough eyes (red ovals depict relative eye size). (C) In the absence of c-cleavage, Gal4 remains tethered at the membrane and therefore is unable to activate Grim expression. This results in flies with increased eye size (i.e., suppression of GAMAREP phenotype). (D) When APP (C99) levels are downregulated, fewer substrate molecules are available (illustrated as one C99 molecule rather than three in B) to activate Grim-dependent apoptosis. This results in larger (suppressed) eye size. (E) When AICD is retained in the cytoplasm, or its levels or function are reduced (not illustrated here), apoptosis occurs less efficiently and the eyes remain large. (F) When AICD-Gal4 mediated transcription is inhibited, apoptosis is also reduced, resulting in flies with larger (suppressed) eyes. In summary, GAMAREP allows the identification of factors that regulate endogenous c-secretase activity (as in C), as well as factors that regulate APP steady state levels (as in D), and AICD function, stability and/or cytoplasmic retention (as in E and F). (G) The false positive reporter. Suppression of GAMAREP can also arise from factors that regulate processes other than APP biology, such as mutations modifying GMR-or Gal4dependent transcriptional activation, or apoptosis, which we collectively named false positives. To eliminate these factors, we employed a ''false positive'' reporter (GMR-Gal4, UAS-apoptotic gene). True modifiers are those that modify they eye phenotypes of GAMAREP, but not those of the false positive reporter. doi:10.1371/journal.pone.0002495.g001 Ubiquilin 1 (UBQLN1) is another gene that has been linked to AD. UBQLN1 encodes a protein with ubiquitin-like (UBL) and ubiquitin-associated (UBA) domains, as well as Sti1 repeats [37], which are often associated with chaperone activity [38]. Several studies suggest links between UBQLN1 and AD. First, in postmortem AD brains, UBQLN1 is found in neurofibrillary tangles [37], a pathological hallmark of AD along with amyloid plaques [1]. Second, the genomic region containing UBQLN1, 9q22, has been identified as containing one major candidate gene for conferring a predisposition to late onset AD [39,40]. Some reports [41,42], but not others [43][44][45][46], suggest that genetic variants in the UBQLN1 gene, including one known as UBQ-8i that deletes one Sti1 repeat, are associated with increased risk for the more prevalent late-onset forms of AD. Further evidence that UBQ-8i has enhanced toxicity comes from studies in Drosophila demonstrating that expression of human UBQ-8i in flies leads to earlier onset and more severe eye degeneration than does expression of wildtype human UBQLN1 [47]. Third, UBQLN1 binds PS1 and PS2 [37], and the fly homolog of Presenilin binds to both human and fly Ubiquilin (Ubqn), respectively [47]. Drosophila ubqn antagonizes presenilin (psn) function in both loss-of-function and gain-of-function studies during development and adult-onset neurodegeneration in vivo [47,48]. Finally, in cultured cells, downregulation of UBQLN1 expression alters APP levels and Ab secretion by modulating APP trafficking [49]. However, reports of altered UBQLN1 expression on APP processing are conflicting [48][49][50]. Because increases in APP levels and/or altered APP processing are likely to be important risk factors for AD, it is crucial to clarify the effect of UBQLN1 on APP in an in vivo system. Additionally, since both X11 and UBQLN1 have been implicated as regulators of APP steady state levels and APP processing, it is crucial to assess the extent to which each molecule contributes to the myriad regulations of APP-related processes, and to determine what their net effect is in vivo.
We engineered transgenic flies that act as living reporters for csecretase activity in the eye [59] (Fig 1B-F). These flies (GMR-C99-Gal4, UAS-grim), known as gamma-secretase activity and APP level reporters (hereafter called GAMAREP), express a chimeric protein which contains a signal sequence, followed by a fusion of the C99 fragment of APP to the yeast transcription factor Gal4. This C99-Gal4 protein is expressed specifically in the developing eye in the presence of a Gal4 responsive element that drives an apoptosis-inducing gene known as grim [62,63]. In the presence of endogenous c-secretase activity (Fig. 1B), AICD-Gal4 is liberated from the membrane, migrates to the nucleus and activates transcription of grim. The resulting apoptosis generates flies with small and rough eyes [59]. When c-secretase activity is impaired (Fig. 1C), Gal4 remains tethered to the membrane, resulting in less apoptosis. Therefore, the size and roughness of the eye inversely correlates, in a semi-quantitative manner, with the level of endogenous c-secretase activity (high levels of activity results in flies with small eyes while low levels are associated with nearly normal eyes). These flies act as sensitive reporters for modest changes (two-fold) in the levels of known c-secretase components [59]. Though originally intended to identify regulators of c-secretase activity [59], the eye phenotypes of GAMAREP flies are also expected to be sensitive to genetic perturbations that alter the levels of APP (Fig. 1D), AICD function, and/or the transcriptional activity of AICD-Gal4 (Fig. 1E, F). To discriminate factors that regulate transgene expression levels, Gal4-dependent transcriptional efficiency or apoptosis, we also generated a false positive reporter (GMR-Gal4, UAS-apoptotic gene) (Fig. 1G). True modifiers will be identified as factors that modify only GAMAREP eye phenotypes, and not those of the false positive reporter. Here we report the identification of X11L and ubqn as regulators of APP in Drosophila.

Increased expression of X11L suppresses GAMAREP eye phenotypes
We carried out a genetic screen (to be described in detail elsewhere) in which GAMAREP flies were crossed with flies carrying single insertions of the EP P element on the X chromosome. EP elements carry a Gal4 responsive promoter pointing outwards from the end of the transposon. Thus, when inserted near the 59 end of a gene, EP elements can drive nearby gene overexpression in a Gal4 dependent manner. From this screen, we identified one suppressor line, which contains an EP element insertion in the 59 region of the X11L gene (CG5675) ( Fig. 2A). To confirm that the GAMAREP suppression is due to overexpression of X11L, we showed that eye-specific overexpression of X11L (GMR-X11L) (Fig. 2E) indeed suppressed the GAMAREP small eye phenotype ( Fig. 2F compared to C). X11L overexpression was confirmed using anti-X11L antibodies (Supplementary Fig. S1). Importantly, X11L overexpression had no effect on the small eye phenotypes of flies expressing apoptosisinducing genes under the direct control of GMR-Gal4 (false positive reporters; [59]) ( Fig. 2G compared to D), indicating that X11L acts specifically on C99, not on GAL4 or apoptotic genes.
To demonstrate that the mechanism we uncovered for X11L is shared by human X11a, we expressed human X11a specifically in the fly eye (Fig. 2H). Strikingly, expression of human X11a also suppressed the small eye phenotype of GAMAREP ( Fig. 2I), but not that of the false positive reporter (Fig. 2J). This suggests that Drosophila X11L and human X11a show functional conservation and that mechanisms of action discovered in flies are likely to be relevant to humans.

X11L overexpression regulates APP at the level of AICD
To determine how increased X11L expression regulates APP in the GAMAREP system, we established assays for APP levels and cleavage in Drosophila. Full length human APP (APP 695 ) or C99 was myc-tagged at the C-terminus and expressed in the eye. Tissue lysates were examined by Western blotting with an anti-Myc antibody to follow the fate of APP (Fig. 3A) or C99 (Fig. 3B). APP-Myc or C99-Myc was cleaved by endogenous aand c-secretase in both systems, yielding an AICD-Myc fragment (c-secretase cleaved) and C83 (an a-secretase cleaved fragment) ( Fig. 3A and 3B). Furthermore, although no endogenous BACE activity has been definitively demonstrated in Drosophila, expression of human BACE in the presence of APP resulted in the generation of C99, in addition to C83 and AICD (Fig. 2C). RNAi-mediated silencing of any one of the four components of the c-secretase complex, psn, nct, aph-1 or pen-2, resulted in attenuated AICD generation and an increase in the levels of CTFs (c-secretase substrates) ( Fig. 3A and  3B). In addition, overexpression of a dominant negative form of psn (Psn D279A which corresponds to human PS1 D257A [59]) resulted in similar effects on APP processing ( Fig. 3C and E).
Since human X11a has previously been shown to bind Presenilin and regulate c-secretase-mediated cleavage in mammalian cells [35], we asked if Drosophila X11L overexpression suppressed GAMAREP by inhibiting c-secretase activity. We crossed flies overexpressing X11L with flies expressing APP-Myc, and their progeny were subjected to Western blot analysis. In contrast to what was observed for the knockdown of psn, nicastrin (nct), aph-1 or pen-2 function, we did not see any reduction in AICD levels or any CTF accumulation ( Fig. 3C and D). Similarly, flies overexpressing X11L in the background of C99-Myc expression also failed to show any evidence of c-secretase inhibition (Fig. 3E). These results suggest that the predominant effect of X11L overexpression in vivo is not due to c-secretase inhibition.
With increased expression of X11L (multiple copies of transgenes), we observed a slight increase in the steady-state levels of APP (data not shown), which is consistent with reported findings in mammals [25,27,33,34]. However, full-length APP and its proteolytic fragments increased largely in proportion. The slight increase in levels of AICD is likely to be physiologically inconsequential, since elevated levels of AICD are predicted to enhance the GAMAREP eye phenotype, whereas we observed a robust suppression.
To explore if X11L overexpression regulates AICD function independently of c-secretase regulation, we generated transgenic flies that overexpressed a fusion protein of the AICD and Gal4 specifically in the eye (GMR-AICD-Gal4). Flies carrying both GMR-AICD-Gal4 and UAS-hid transgenes ( Fig. 4A and 4E, hereafter called the AICDREP) were viable but exhibited an almost complete absence of eye tissue, due to extensive apoptosis of the cells in the eye [62][63][64]. This AICDREP (Fig. 4E) displayed a much smaller eye size than that of GAMAREP (Fig. 2C), presumably because in the AICDREP, Gal4 is not tethered to the membrane as compared to C99-Gal4, and therefore does not require any liberation by c-secretase. The AICDREP thus serves as a useful system to identify genes that regulate AICD function in vivo ( Fig. 4B-D). Remarkably, overexpression of X11L in the presence of AICDREP almost completely restored the eye size to wildtype (Fig. 4F). This suggests that the major effect of X11L overexpression is the c-secretase independent inhibition of AICD. In addition, expression of human X11a suppressed the AICDREP as effectively as overexpression of X11L (Fig. 4H), indicating functional conservation between Drosophila and human X11.
Mammalian X11a has been shown to bind to the AICD via its PTB domain [30,31]. If X11L overexpression exerts its major effect on AICD, we would expect that expression of a truncated X11L lacking its PTB domain (X11LDPTB) should no longer be able to suppress the GAMAREP or AICDREP phenotypes. Indeed, eye-specific overexpression of X11LDPTB (Fig. 2K) failed to suppress both GAMAREP and the AICDREP (Fig. 2L and 4G), further supporting the hypothesis that X11L regulates AICD function. Importantly, neither overexpression of X11L, X11LDPTB nor human X11a affected the GMR-Gal4, UAShid)\ eye phenotype (Fig. 4I-L), suggesting that X11 expression acts specifically on AICD. Furthermore, the expression of X11L or X11a did not affect AICD steady-state levels (Fig. 4M), suggesting that X11 may promote cytoplasmic retention of AICD, or inhibit its ability to regulate transcription.
X11L overexpression does not affect the c-secretase mediated cleavage of Notch Our observation that overexpression of fly and human X11 does not alter c-secretase mediated cleavage of APP stands in contrast to a previous report [36]. To explore this issue further, we asked Overexpression of X11L (GMR-X11L) shows no external eye phenotypes (E), but strongly suppresses GAMAREP eye phenotypes (F). Expression of human X11a (GMR-X11a), which by itself does not produce any visible eye phenotypes (H), suppresses GAMAREP phenotypes as effectively as fly X11L overexpression (I). X11L's ability to suppress GAMAREP is dependent on its PTB domain, since overexpression of X11LDPTB (GMR-X11LDPTB) (K) fails to suppress GAMAREP eye phenotype (L). The ability of X11 to suppress GAMAREP eye phenotype is specific, since overexpression of X11L, X11DPTB or human X11 fails to suppress the eye phenotype of the false-positive reporter (GMR-Gal4, UAS-grim/reaper) (D, G, J and M). The X11LDPTB transgene expresses a truncated protein as detected by anti-X11L antibodies [75]. Of note, strong overexpression of X11L (multiple copies of GMR-X11L or GMR-Gal4, UAS-X11L) can result in a rough eye phenotype (data not shown), consistent with previous reports [76]. doi:10.1371/journal.pone.0002495.g002 whether X11L overexpression could regulate c-cleavage of Notch, another substrate of c-secretase. Notch is a transmembrane receptor that mediates cell-cell communication events in multiple developmental contexts [65,66]. Following cleavage of its extracellular domain (the S2 cleavage), Notch requires csecretase-mediated cleavage (denoted S3) to release the Notch intracellular domain (NICD), and activate downstream signaling events [67]. To analyze Notch processing, a transmembrane version of Notch representing the S2 cleaved fragment was Myctagged at its C-terminus and specifically expressed in the fly eye. Silencing of any one of the four c-secretase components led to decreased c-cleavage of Notch, as demonstrated by decreased levels of NICD, and an increase in the levels of uncleaved Notch (Fig. 5A). Overexpression of Psn D279A also caused reduced levels of NICD and an accumulation of uncleaved Notch (Fig. 5B). In contrast, neither X11L overexpression nor expression of human X11a led to any significant effects on NICD production (Fig. 5B).
To further test the hypothesis that overexpression of X11L does not affect c-secretase activity, we searched for a genetic interactions between X11L and psn in vivo. RNAi-mediated silencing of psn specifically in the eye (GMR-RNAi-psn) resulted in flies with small, rough eyes (Fig. 5C). This is likely due to impaired cleavage of Notch since partial loss of Notch function also results in small, rough eyes [68]. As expected, overexpression of psn partially suppressed the rough eye phenotype associated with GMR-RNAipsn, while expression of Psn D279A enhanced this phenotype (data not shown). In contrast, expression of either Drosophila X11L or human X11 did not result in enhancement of the GMR-RNAi-psn eye phenotype (Fig. 5D and 5E), suggesting that overexpression of X11L does not significantly inhibit c-secretase activity in vivo.
ubqn acts to stabilize the steady-state levels of APP To study ubqn function in Drosophila, we generated two RNAi constructs targeted to two independent regions of ubqn (the coding region and the 39-untranslated region, respectively), and used these transgenic flies to carry out tissue-specific silencing [47]. As described previously, ubiquitous expression of RNAi-ubqn resulted in a loss of detectable Ubqn protein [47]. Both ubqn RNAi In contrast, expression of neither X11L nor human X11a results in any alteration in the CTF levels. (D) Western blot quantifications of APP and its fragments suggest that there are no significant differences in the levels of either full length APP or its fragments as compared to control. Values are normalized to the corresponding tubulin loading controls. Data from X11 overexpression (dark bars) and control (white bars) are analyzed by paired t-test. (E) When C99-9Myc is expressed, expression of Psn DN results in a decrease in AICD levels and an increase in C83 levels, while overexpression of X11L fails to show any changes. Note that panels depicting CTFs are from the same Western blots exposed for different lengths of time than the corresponding APP panels. For each sample, sibling flies of the appropriate genotype raised under identical conditions were used as controls, and are shown immediately adjacent to the corresponding experimental sample (linked by a horizontal bar). doi:10.1371/journal.pone.0002495.g003 transgenes silenced ubqn expression to similar levels and gave identical phenotypes in all experiments [47]. Using the same RNAi hairpins, we carried out eye-specific silencing of ubqn (GMR-RNAi-ubqn); effects confirmed by Western blotting of head lysates using anti-Ubqn antibodies (Supplementary Fig. S1). Though silencing of ubqn function did not lead to any overt developmental phenotypes (Fig. 6C), it strongly suppressed the GAMAREP phenotype ( Fig. 6B compared with 6A). Importantly, silencing of ubqn did not alter the false positive reporter eye phenotypes ( Fig. 6D compared with 2D), suggesting that the suppression of GAMAREP was specifically due to an effect on C99.
We next investigated how ubqn loss-of-function suppressed GAMAREP. APP-Myc was co-expressed with human BACE in the eye. Silencing of ubqn resulted in a reduction of steady state levels of full length APP as well as its proteolytic fragments (C99, C83 and AICD) in the eye (Fig. 6G). Importantly, the reduction of full length APP and its cleaved forms occurred largely propor- tionately (Fig. 6G), suggesting that loss of ubqn leads to a decrease in the overall steady-state levels of APP, not a decrease in the activity of any particular secretase.
Since silencing of ubqn resulted in a decrease in AICD levels as a consequence of reduced APP levels, we asked if silencing ubqn would also affect the AICD reporter eye phenotypes. RNAi-ubqn flies were crossed to flies overexpressing Myc-tagged AICD. Indeed, we observed that eye-specific expression of RNAi-ubqn led to a modest suppression of the AICDREP (Fig. 6F compared with  6E), and a reduction in AICD levels in vivo (Fig. 6H).
To determine if ubqn acts as a dose-dependent regulator of APP levels, we examined the consequences of ubqn overexpression on APP levels. Eye-specific overexpression of ubqn (GMR-ubqn) did not generate any overt eye phenotypes, or suppress the GAMA-REP eye phenotypes (data not shown). Expression levels of ubqn overexpression was confirmed by anti-Ubqn antibodies (Supple-mentary Fig. S1). However, ubqn overexpression (GMR-Gal4, UAS-ubqn) led to an increase in APP levels (Fig. 6I). As with silencing of ubqn, ubqn overexpression increased the steady state levels of the full length and c-secretase cleaved fragments of APP largely in proportion, suggesting that perturbation of Ubqn levels does not affect c-cleavage of APP (Fig. 6I). Together, we concluded that the suppression of the GAMAREP phenotype caused by loss of ubqn function is most likely the result of a decrease in APP steady-state levels. While loss of ubqn strongly suppressed the GAMAREP phenotype, we did not observe a significant enhancement of GAMAREP upon ubqn overexpression. This is most likely due to the fact that our GAMAREP reporter is much less sensitive in detecting enhancers than suppressors. This is based on multiple screens carried out to identify modifiers of GAMA-REP, from which we isolated a dozen GAMAREP suppressors but no enhancers (unpublished observations).
How Ubqn regulates APP levels is unclear, but could involve direct interactions between these proteins. To explore this hypothesis, we asked if Ubqn can bind to APP directly. We mixed purified GST and GST-Ubqn fusion proteins with lysates from Schneider 2 (S2) cells transfected with Myc-tagged C99. Indeed, Myc-C99 specifically bound GST-Ubqn, but not GST alone (Fig. 6J). Interestingly, in addition to native-sized C99, GST-Ubqn preferentially bound to high molecular weight products, which are likely polyubiquitinated forms of C99 (Fig. 6J). Deletion of the UBA domain, but not the UBL domain, significantly abrogated the interaction between Ubqn and C99 (Fig. 6J), indicating that the UBA domain is required for this interaction.

Discussion
Here, we report the identification of two APP regulators, X11L and Ubqn, using a living reporter, GAMAREP. GAMAREP allows us to effectively identify regulators of c-secretase activity, APP steady-state levels and/or AICD function. Both the steadystate levels and the processing of APP have been demonstrated to be important in the pathogenesis of AD. However, the function of AICD, a much less studied APP fragment, may also be crucial for AD pathogenesis, since AICD is generated along with Ab and may have functions including transcriptional regulation, signaling, apoptosis and calcium homeostasis. Importantly, when a protein regulates multiple aspects of APP, either directly or indirectly, the combination of GAMAREP, AICDREP, and our in vivo cleavage assays allows us to uncover the relative contributions of these regulatory inputs. This is of particular relevance in this study since both X11 and UBQLN1 have been proposed to have multiple functions affecting myriad pathways in various systems. Therefore, it is important to determine in vivo which pathways play predominant roles in events related to AD pathogenesis. Our data suggest that the predominant role of X11L is to regulate APP at the level of the AICD by a process requiring its PTB domain, whereas the major role of Ubqn is to regulate APP levels, likely through a direct physical interaction with APP that is dependent on the UBA domain of Ubqn.
As a modifier of AICDREP, X11L does not regulate AICD levels. Following c-cleavage, AICD is thought to migrate into the nucleus. It is likely that X11L overexpression causes cytoplasmic retention of AICD based on the following observations. In mammalian cultured cells, X11 overexpression can reduce nuclear localization of the AICD with Tip60 [69]. Moreover, X11 can shuttle between the nucleus and cytoplasm to regulate AICD transcriptional targets [70]. Our functional studies using AIC-DREP suggest that the PTB domain of X11L is essential for X11's ability to regulate AICD. Therefore, our hypothesis is that X11L overexpression leads to the cytoplasmic retention of AICD via a direct physical interaction between the PTB domain of X11L and AICD.
We have also showed that, in contrast to prior studies [36], X11L overexpression does not inhibit c-secretase activity in vivo. In Drosophila, none of the APP cleavage products, C83, C99 or AICD, is affected by X11L overexpression. Moreover, X11L overexpression does not inhibit c-secretase activity in vivo towards another key substrate, Notch, at the level of cleavage, or at the level of genetic interaction with psn. In a prior study, King et al concluded that X11 inhibits c-secretase activity, based largely on the findings that extracellular Ab levels were decreased upon X11 overexpression [36]. They interpreted this to be due to decreased Ab production by c-secretase inhibition. However, decreased extracellular Ab levels could also result from decreased Ab secretion or increased Ab degradation. In fact, another in vivo study suggested that although X11 overexpression led to reduced Ab levels, levels of C83 and C99 (direct c-secretase substrates) did not increase [29]. Since overexpression of X11 can increase the steady-state levels of APP, likely by modulating APP trafficking to certain compartments in the secretory pathway [25,27,33,34], an alternative interpretation of work by King et al is that X11 overexpression decreases Ab via its modulation of APP trafficking, rather than by regulating c-secretase cleavage. Therefore, these prior studies are, in fact, consistent with our data and interpretation.
Our data also suggest that the predominant function of ubqn is to regulate APP levels, not c-secretase activity. We have established that Ubqn can bind to and stabilize APP. In contrast to the selective reduction of AICD and increase in CTFs seen following the knockdown of any one of the four c-secretase components, both ubqn loss-of-function and ubqn overexpression modify the levels of APP and its cleavage fragments largely in proportion. Our results are consistent with findings made by Zhang et al and different from Hiltunen et al; possible reasons for these differences are discussed in Zhang et al [50].
In this study, we did not observe any alterations in c-secretase activity on APP in response to the silencing of ubqn or ubqn overexpression, similar to observations made by others [48][49][50]. However, we have previously shown that Ubqn binds to Psn and antagonizes psn function both during development and during adult-onset eye neurodegeneration in Drosophila [47]. There are several ways these observations can be reconciled. c-secretase activity is known to vary depending upon the substrate and the tissue in which it acts [71]. It is possible that ubqn antagonizes psn activity such that its effect on c-secretase is more evident on substrates other than APP in the eye. As a result, the modification of the c-cleavage of APP by ubqn, if any, might be below the detection limits of our assay. Alternatively, it is well established that in addition to being the catalytic core of the c-secretase complex, Psn can also function in a c-secretase independent fashion [72,73]. Therefore, it is possible that ubqn primarily inhibits c-secretase independent functions of psn, leaving csecretase dependent activity largely intact. In either case, the primary effect of ubqn on APP appears to be the regulation of APP steady state levels, and not inhibition of the c-cleavage of APP.
Here, using GAMAREP and AICDREP, in conjunction with in vivo cleavage analysis, we have identified factors that regulate APP. Further screens using these tools are likely to identify other proteins which may have implications for AD pathogenesis. Since increased expression of X11 attenuates Ab production or secretion in APP transgenic mice [28,29], but does not act on c-secretase directly (this work), X11 may present an attractive therapeutic target for AD. Similarly, since reduced Ubqn levels result in a modest decrease in APP levels, whereas expression of both human wildtype and the AD variant of UBQLN1 leads to adult onset, age-dependent eye degeneration in Drosophila [47], UBQLN1 may also be considered as a possible drug target for AD.

Molecular Biology
A microRNA-based technology [74] was used for RNAi silencing. To silence psn, nct, aph-1 and pen-2, the respective coding regions were independently targeted. PCR products of these microRNA precursors were cloned into pGMR. To generate GMR-X11L and UAS-X11L, the X11L coding sequence was PCR amplified from the EST clone, LD29081, and subcloned into each vector. To generate GMR-human-X11a, a clone of X11a (kindly provided by Declan McLoughlin) was subcloned into the EcoR1 and Not1 sites of pGMR. To generate GMR-X11DPTB, the insert of UAS-X11LDPTB [75] was subcloned into pGMR vector. To generate GMR-AICD-Gal4, the nucleotide region encoding the predicted human AICD fragment was PCR amplified and fused in-frame, upstream of an S. cerevisiae Gal4 sequence. To make GMR-ubqn, GMR-RNAi-ubqnCDS and GMR-RNAi-ubqnUTR, the inserts from UAS-ubqn, UAS-RNAi-ubqnCDS and UAS-RNAi-ubqnUTR [47] were subcloned into pGMR vector, respectively. To generate pGMR-APP-9Myc and GMR-AICD-9Myc, coding sequence of APP 695 or AICD was PCR amplified and cloned inframe upstream of a 9xMyc sequences that has been previously subcloned into pGMR vector. C99-9Myc was also generated by PCR and subcloned to generate pMT-C99-9Myc and GMR-C99-9Myc, respectively. To make pGEX-UbqnDUBL, a PCR fragment encoding UbqnDUBL was subcloned into the EcoR1 and Not1 sites of the modified pGEX4T-1 vector. pGEX-Ubqn and pGEX-UbqnDUBA, were described previously [47]. All cloned PCR products were confirmed by DNA sequencing.

Drosophila Genetics and Strains
For the X chromosome EP screen, individual lines of females carrying a single inserted EP element were individually crossed to GAMAREP males. Progeny carrying one copy of an EP insertion and one copy of GAMAREP were examined, and their eye size compared with control flies derived from a cross of w 1118 females to GAMAREP males. For experiments involving transgenic flies, multiple independent fly lines were generated (Rainbow Transgenic Flies) and tested for each transgene. UAS-BACE flies [60] were obtained from Rita Reifegerste via Doris Kretzschmar, and UAS-hid flies [64] were obtained from Bruce Hay. GMR-Psn D257A was described previously [59].

Scanning Electron Microscopy
Freshly sacrificed flies were mounted on their side with one eye upward on white tape using clear nail polish. All flies were placed on a rotating platform to permit for orientation under vacuum and were imaged at 1806 magnification and 100 psi using a Hitachi 2460N scanning electron microscope. Analysis of eye phenotypes was performed as described previously [59].

Antibody Generation
A fusion of GST-TEV to Drosophila X11L residues 452-775, corresponding to a region within the N-terminal domain, with an intervening TEV protease recognition site, was purified from E. coli lysates. Glutathione agarose-retained proteins were cleaved by TEV protease to remove GST, and used to immunize rabbits (Imgenex).

Lysate Preparation and Western Blotting
Heads from age and sex-matched adults were disrupted in lysis buffer, complete protease inhibitor cocktail (Roche) using a sonicator-3000 from MISONIX. Samples were sonicated. Samples were boiled, centrifuged, and total protein from 4 heads per genotype was analyzed by Western blotting. Antibodies used were anti-Myc (Upstate), anti-Ubqn [47], anti-X11L and anti-Tubulin (Sigma).

S2 Cell Culture and Transfection
S2 cells were grown in Schneider's Drosophila medium (Invitrogen) supplemented with 10% FBS, 50 units/ml penicillin, and 50 mg/ml streptomycin at room temperature. Transfections were carried out using MaxFect transfection reagent (Molecula). Typically, 1.5610 6 cells plated in a 12-well dish were transfected with 0.7 -1 mg total plasmid DNA plus 5 mL MaxFect reagent. Metallothionein promoter expression was induced with 0.5 mM copper sulfate 24 hours after transfection. S2 cell lysates were prepared by harvesting transfected cells (,5610 6 ) in 1 mL lysis buffer, followed by incubation on ice for 10 minutes and centrifugation (16,0006g, 10 min) to pellet insoluble debris.

GST Pulldown Assay
10 mg GST fusion proteins purified from E. coli lysates were retained on 15 mL glutathione beads, and mixed with 400 mg S2 cell lysate in 1mL total volume. Retained proteins were eluted by boiling in Laemmli sample buffer, and detected by Western blotting with anti-Myc antibody (Covance) [47]. Figure S1 The expression levels of X11L and ubqn transgenes. (A) We have generated polyclonal antibodies against X11L. X11L overexpression is accomplished using either an eye-specific promoter (GMR-X11L) or the eye-specific driver using the UAS-GAL4 system (GMR-Gal4, UAS-X11L). Western blots of lysates from fly heads overexpressing X11L or lysates from Schneider 2 cells overexpressing X11L (MT-X11L) reveal a band of predicted size using anti-X11L antibodies. (B) Western blots of lysates from fly heads expressing RNAi-ubqn or ubqn using anti-Ubqn antibodies. Silencing of ubqn significantly reduces Ubqn levels, while ubqn overexpression increases Ubqn levels. A nonspecific band (*) serves as protein loading control. Found at: doi:10.1371/journal.pone.0002495.s001 (1.06 MB TIF)